A Research On Data Offloading Methods Based On Personal Mobility Patterns

In recent years,with the continuous progress in mobile communication technology and the rapid growth of the amount of mobile devices,global mobile data traffic has increased sharply.This has a great impact on mobile users and communication operators.For mobile users,mobile communication networks can be easily congested by large-scale mobile data access requests,which degrades users experience seriously;For communication operators,in order to meet the sharply increasing users'mobile data access requests,and alleviate the pressure on mobile communication networks,they need to constantly purchase spectrum,build more base stations and upgrade networks,which is a considerable expense.How to alleviate the pressure on mobile communication networks,meet the increasing mobile data requests of users and ensure users experience has become an urgent problem for communication operators.To cope with the challenges brought by the rapid growth of mobile data traffic,data originally transmitted via mobile communication network could be transmitted through other networks leveraging data offloading methods.It has been proved that data offloading methods can alleviate the pressure on mobile communication network and meet users'mobile data access requests efficiently.In addition,it is found that considering personal mobility in data offloading methods can improve performance of offloading methods.Therefore,a research on data offloading methods based on personal mobility patterns is conducted.It mainly includes the following aspects:(1)This paper studies the methods to mine patterns of personal mobility.Firstly,personal mobility is analyzed and two patterns of personal mobility are concluded:1)Users stay in a few hotspot positions(hereinafter referred to as hotspots)most of the day;2)Users frequently move between hotspots in a day-to-day cycle and prefer to move along frequent paths.Then,the Kalman filtering method is used to preprocess users' trajectories,considering the ellipsometric error of the GPS device and the weak GPS signal in some regions.Finally,a hotspots mining method and a frequent paths mining method are proposed to mine above two patterns respectively for subsequent data offloading methods.(2)The progress of mobile communication technology has brought streaming media services into reality.It has become a fashion to watch videos on mobile devices.In fact,mobile video has the characteristics of personalization and cacheability.To this end,a personal hotspot positions based data offloading method is proposed(hereinafter referred to as HPO).HPO includes three steps:1)Mining personal hotspots;2)Based on the history of users'videos viewing,HPO models users'interests in mobile videos by jointly leveraging method ItemCF/CB and GBDT+LR,and then generates the most interesting videos for users;3)When users are at hotspots and the mobile devices are connected to the complementary network,HPO will buffer the mobile videos into users'devices to offload mobile video.Finally,the performance of HPO is verified on MovieLens dataset by comparing with GBDT+LR,GBDT and LR.The experimental results show that HPO has higher buffering accuracy and execution efficiency.(3)A fact should not be ignored that some data is not cacheable and these data accounts for a large proportion of mobile data traffic.Since these data cannot be offloaded in pre-caching fashion,there is still a need to develop an appropriate offloading method.Given that there are a large number of complementary networks deployed in urban areas and their distribution information is available,a personal frequent paths based data offloading method is proposed(hereinafter referred to as FPO).For each user,FPO mines his/her frequent paths firstly.Then,based on the distribution information of complementary networks,the coverage area of every complementary network along the frequent path is determined.After that,the time user takes to enter/exit every complementary network coverage is predicted by using the spatio-temporal patterns of personal mobility on frequent path;Finally,when user is moving along the frequent path,FPO makes an optimal data offloading strategy according to deadlines of all transmission tasks,so as to make full use of all available complementary networks.Extensive experiments are conducted to verify the performance of FPO with the deterministic delay offloading method and the adaptive delay offloading method selected as benchmark methods.The experimental results show that FPO could improve the efficiency of data offloading and reduce data transmission time obviously.